Simulating the effect of tillage practices with the global ......(Levis et al., 2014; Oleson et al., 2010) PROMET (Mauser and Bach, 2009), and the Dynamic Land Ecosystem Model (DLEM;

Geosci. Model Dev., 12, 2419–2440, 2019https://doi.org/10.5194/gmd-12-2419-2019© Author(s) 2019. This work is distributed underthe Creative Commons Attribution 4.0 License.

Simulating the effect of tillage practices with the globalecosystem model LPJmL (version 5.0-tillage)Femke Lutz1,2, Tobias Herzfeld1, Jens Heinke1, Susanne Rolinski1, Sibyll Schaphoff1, Werner von Bloh1,Jetse J. Stoorvogel2, and Christoph Müller11Potsdam Institute for Climate Impact Research (PIK), member of the Leibniz Association,P.O. Box 60 12 03, 14412 Potsdam, Germany2Wageningen University, Soil Geography and Landscape Group, P.O. Box 47, 6700 AA Wageningen, the Netherlands

Correspondence: Femke Lutz ([email protected])

Received: 12 October 2018 – Discussion started: 13 November 2018Revised: 26 April 2019 – Accepted: 15 May 2019 – Published: 19 June 2019

Abstract. The effects of tillage on soil properties, crop pro-ductivity, and global greenhouse gas emissions have beendiscussed in the last decades. Global ecosystem models havelimited capacity to simulate the various effects of tillage.With respect to the decomposition of soil organic matter,they either assume a constant increase due to tillage or theyignore the effects of tillage. Hence, they do not allow foranalysing the effects of tillage and cannot evaluate, for ex-ample, reduced tillage or no tillage (referred to here as “no-till”) practises as mitigation practices for climate change. Inthis paper, we describe the implementation of tillage-relatedpractices in the global ecosystem model LPJmL. The ex-tended model is evaluated against reported differences be-tween tillage and no-till management on several soil proper-ties. To this end, simulation results are compared with pub-lished meta-analyses on tillage effects. In general, the modelis able to reproduce observed tillage effects on global, as wellas regional, patterns of carbon and water fluxes. However,modelled N fluxes deviate from the literature values and needfurther study. The addition of the tillage module to LPJmL5opens up opportunities to assess the impact of agriculturalsoil management practices under different scenarios with im-plications for agricultural productivity, carbon sequestration,greenhouse gas emissions, and other environmental indica-tors.

1 Introduction

Agricultural fields are tilled for various purposes, includingseedbed preparation, incorporation of residues and fertiliz-ers, water management, and weed control. Tillage effectsa variety of biophysical processes that affect the environ-ment, such as greenhouse gas emissions or soil carbon se-questration and can influence various forms of soil degrada-tion (e.g. wind, water, and tillage erosion) (Armand et al.,2009; Govers et al., 1994; Holland, 2004). Reduced tillageor no tillage (hereafter referred to as “no-till”) is being pro-moted as a strategy to mitigate greenhouse gas (GHG) emis-sions in the agricultural sector (Six et al., 2004; Smith etal., 2008). However, there is an ongoing long-lasting de-bate about tillage and no-till effects on soil organic carbon(SOC) and GHG emissions (e.g. Lugato et al., 2018). In gen-eral, reduced tillage and no-till tend to increase SOC storagethrough a reduced decomposition and consequently reducesGHG emissions (Chen et al., 2009; Willekens et al., 2014).However, discrepancies exist on the effectiveness of reducedtillage or no-till on GHG emissions. For instance, Abdalla etal. (2016) found in a meta-analyses that on average no-tillsystems reduce CO2 emissions by 21 % compared to con-ventional tillage, whereas Oorts et al. (2007) found that CO2emissions from no-till systems increased by 13 % comparedto conventional tillage, and Aslam et al. (2000) found onlyminor differences in CO2 emissions. These discrepancies arenot surprising as tillage effects a complex set of biophysicalfactors, such as soil moisture and soil temperature (Snyder etal., 2009), which drive several soil processes, including thecarbon and nitrogen dynamics and crop performance. More-

Published by Copernicus Publications on behalf of the European Geosciences Union.

2420 F. Lutz et al.: Simulating the effects of tillage

over, other factors such as management practices (e.g. fertil-izer application and residue management) and climatic con-ditions have been shown to be important confounding factors(Abdalla et al., 2016; Oorts et al., 2007; van Kessel et al.,2013). For instance, Oorts et al. (2007) attributed the higherCO2 emissions under no-till to higher soil moisture and de-composition of crop litter on top of the soil. Van Kessel etal. (2013) found that N2O emissions were smaller under no-till in dry climates and that the depth of fertilizer applicationwas important. Finally, Abdalla et al. (2016) found that no-till effects on CO2 emissions are most effective in drylandsoils.

In order to upscale this complexity and to study the role oftillage for global biogeochemical cycles, crop performance,and mitigation practices, the effects of tillage on soil prop-erties need to be represented in global ecosystem models.Although tillage is already implemented in other ecosystemmodels at different levels of complexity (Lutz et al., 2019;Maharjan et al., 2018), tillage practices are currently under-represented in global ecosystem models that are used for bio-geochemical assessments. In these, the effects of tillage areeither ignored or represented by a simple scaling factor ofdecomposition rates. Global ecosystem models that ignorethe effects of tillage include, for example, JULES (Best etal., 2011; Clark et al., 2011), the Community Land Model(Levis et al., 2014; Oleson et al., 2010) PROMET (Mauserand Bach, 2009), and the Dynamic Land Ecosystem Model(DLEM; Tian et al., 2010). The models in which the effectsof tillage are represented as an increase in decomposition in-clude LPJ-GUESS (Olin et al., 2015; Pugh et al., 2015) andORCHIDEE-STICS (Ciais et al., 2011).

The objective of this paper is to (1) extend the Lund Pots-dam Jena managed Land (LPJmL5) model (von Bloh et al.,2018) so that the effects of tillage on biophysical processesand global biogeochemistry can be represented and studiedand (2) evaluate the extended model against data reported inmeta-analyses by using a set of stylized management scenar-ios. This extended model version allows for quantifying theeffects of different tillage practices on biogeochemical cy-cles, crop performance, and for assessing questions related toagricultural mitigation practices. Despite uncertainties in theformalization and parameterization of processes, the process-based representation allows for enhancing our understandingof the complex response patterns as individual effects andfeedbacks can be isolated or disabled to understand their im-portance. To our knowledge, some crop models that havebeen used at the global scale, e.g. EPIC (Williams et al.,1983) and DSSAT (White et al., 2010), have similarly de-tailed representations of tillage practices but models used tostudy the global biogeochemistry (Friend et al., 2014) haveno or only very coarse representations of tillage effects.

2 Tillage effects on soil processes

Tillage affects different soil properties and soil processes, re-sulting in a complex system with various feedbacks on pro-cesses related to soil water, temperature, carbon (C), and ni-trogen (N) (Fig. 1). The effect of tillage has to be imple-mented and analysed in conjunction with residue manage-ment as these management practices are often interrelated(Guérif et al., 2001; Strudley et al., 2008). The processesthat were implemented into the model were chosen based onthe importance of the process and its compatibility with theimplementation of other processes within the model. Thoseprocesses are visualized in Fig. 1 with solid lines; processesthat have been ignored in this implementation are visualizedwith dotted lines. To illustrate the complexity, we here de-scribe selected processes in the model affected by tillage andresidue management, using the numbered lines in Fig. 1.

With tillage, surface litter is incorporated into the soil (1)and increases the soil organic matter (SOM) content of thetilled soil layer (2) (Guérif et al., 2001; White et al., 2010),while tillage also decreases the bulk density of this layer (3)(Green et al., 2003). An increase in SOM positively affectsthe porosity (4) and therefore the soil water holding capac-ity (whc) (5) (Minasny and McBratney, 2018). Tillage alsoaffects the whc by increasing porosity (6) (Glab and Kulig,2008). A change in whc affects several water-related pro-cesses through soil moisture (7). For instance, changes insoil moisture influence lateral runoff (8) and leaching (9)and affect infiltration. A wet (saturated) soil, for example,decreases infiltration (10), while infiltration can be enhancedif the soil is dry (Brady and Weil, 2008). Soil moisture af-fects primary production as it determines the amount of wa-ter which is available for the plants (11) and changes in plantproductivity again determine the amount of residues left atthe soil surface or to be incorporated into the soil (1) (feed-back not shown).

The presence of crop residues on top of the soil (referred toas “surface litter” hereafter) enhances water infiltration intothe soil (12) (Guérif et al., 2001; Jägermeyr et al., 2016;Ranaivoson et al., 2017), and thus increases soil moisture(13). That is because surface litter limits soil crusting, canconstitute preferential pathways for water fluxes, and slowslateral water fluxes at the soil surface so that water has moretime to infiltrate (Glab and Kulig, 2008). Consequently, sur-face litter reduces surface runoff (14) (Ranaivoson et al.,2017). Surface litter also intercepts part of the rainfall (15),reducing the amount of water reaching the soil surface, butalso lowers soil evaporation (16) and thus reduces unproduc-tive water losses to the atmosphere (Lal, 2008; Ranaivoson etal., 2017). Surface litter also reduces the amplitude of varia-tions in soil temperature (17) (Enrique et al., 1999; Steinbachand Alvarez, 2006). The soil temperature is strongly relatedto soil moisture (18), through the heat capacity of the soil, i.e.a relatively wet soil heats up much slower than a relativelydry soil (Hillel, 2004). The rate of SOM mineralization is

Geosci. Model Dev., 12, 2419–2440, 2019 www.geosci-model-dev.net/12/2419/2019/

F. Lutz et al.: Simulating the effects of tillage 2421

Figure 1. Flow chart diagram of feedback processes caused by tillage, which are considered (solid lines) and not considered (dashed lines)in this implementation in LPJmL5.0-tillage. Blue lines highlight positive feedbacks, red negative, and black are ambiguous feedbacks. Thenumbers in the figure indicate the processes described in Sect. 2.

influenced by changes in soil moisture (19) and soil temper-ature (20) (Brady and Weil, 2008). The rate of mineralizationaffects the amount of CO2 emitted from soils (21) and the in-organic N content of the soil. Inorganic N can then be takenup by plants (22), be lost as gaseous N (23), or transformedinto other forms of N. The processes of nitrate (NO−3 ) leach-ing, nitrification, denitrification, mineralization of SOM, andimmobilization of mineral N forms are explicitly representedin the model (von Bloh et al., 2018). The degree to whichsoil properties and processes are affected by tillage mainlydepends on the tillage intensity, which is a combination oftillage efficiency and mixing efficiency (explained in detailin Sect. 3.2 and 3.5.2). Tillage has a direct effect on the bulkdensity of the tilled soil layer. The type of tillage determinesthe mixing efficiency, which affects the amount of incorpo-rating residues into the soil. Over time, soil properties recon-solidate after tillage, eventually returning to pre-tillage states.The speed of reconsolidation depends on soil texture and thekinetic energy of precipitation (Horton et al., 2016).

This implementation mainly focuses on two processes di-rectly affected by tillage: (1) the incorporation of surface lit-ter associated with tillage management and subsequent ef-fects (Fig. 1, path 1 and following paths) and (2) the decrease

in bulk density and the subsequent effects of changed soilwater properties (Fig. 1, e.g. path 3 and following paths).In order to limit model complexity and associated uncer-tainty, tillage effects that are not directly compatible with theoriginal model structure (such as subsoil compaction) or re-quire very high spatial resolution are not taken into accountin this initial tillage implementation, despite acknowledgingthat these processes can be important.

3 Implementation of tillage routines into LPJmL

3.1 LPJmL model description

The tillage implementation described in this paper was in-troduced into the dynamical global vegetation, hydrology,and crop-growth model LPJmL. This model was recently ex-tended to also cover the terrestrial N cycle, accounting forN dynamics in soils and plants and N limitation of plantgrowth (LPJmL5; von Bloh et al., 2018). Previous compre-hensive model descriptions and developments are describedby Schaphoff et al. (2018a). The LPJmL model simulatesthe C, N, and water cycles by explicitly representing bio-physical processes in plants (e.g. photosynthesis) and soils

www.geosci-model-dev.net/12/2419/2019/ Geosci. Model Dev., 12, 2419–2440, 2019


(e.g. mineralization of N and C). The water cycle is repre-sented by the processes of rain water interception, soil andlake evaporation, plant transpiration, soil infiltration, lateraland surface runoff, percolation, seepage, routing of dischargethrough rivers, storage in dams and reservoirs, and water ex-traction for irrigation and other consumptive uses.

In LPJmL5, all organic matter pools (vegetation, litter, andsoil) are represented as C pools and the corresponding Npools with variable C : N ratios. Carbon, water, and N poolsin vegetation and soils are updated daily as the result of com-puted processes (e.g. photosynthesis, autotrophic respiration,growth, transpiration, evaporation, infiltration, percolation,mineralization, nitrification, and leaching; see von Bloh etal., 2018, for the full description). Litter pools are repre-sented by the aboveground pool (e.g. crop residues, such asleaves and stubble) and the belowground pool (roots). Thelitter pools are subject to decomposition, after which the hu-mified products are transferred to the two SOM pools thathave different decomposition rates (Fig. S1a in the Supple-ment). The fraction of litter which is harvested from the fieldcan range between almost fully harvested or not harvested,when all litter is left on the field (90 %; Bondeau et al., 2007).In the soil, pools of inorganic, reactive N forms (NH+4 , NO

−

3 )are also considered. Each organic soil pool consists of C andN pools and the resulting C : N ratios are flexible. Soil C : Nratios are considerably smaller than those of plants as im-mobilization by microorganisms concentrates N in SOM. InLPJmL, a soil C : N ratio of 15 is targeted by immobilizationfor all soil types (von Bloh et al., 2018). The SOM pools inthe soil consist of a fast pool with a turnover time of 30 years,and a slow pool with a 1000-year turnover time (Schaphoffet al., 2018a). Soils in LPJmL5 are represented by five hy-drologically active layers, each with a distinct layer thick-ness. The first soil layer, which is mostly affected by tillage,is 0.2 m thick. The following soil layers are 0.3, 0.5, 1.0, and1.0 m thick, followed by a 10.0 m bedrock layer, which servesas a heat reservoir in the computation of soil temperatures(Schaphoff et al., 2013).

LPJmL5 has been evaluated extensively and demonstratedgood skills in reproducing C, water, and N fluxes in both agri-cultural and natural vegetation on various scales (von Bloh etal., 2018; Schaphoff et al., 2018b).

3.2 Litter pools and decomposition

In order to address the residue management effects of tillage,the original aboveground litter pool is now separated into anincorporated litter pool (Clitter,inc) and a surface litter pool(Clitter,surf) for carbon, and the corresponding pools (Nlitter,incand Nlitter,surf) for nitrogen (Fig. S1b in the Supplement).Crop residues not collected from the field are transferred tothe surface litter pools. A fraction of residues from the sur-face litter pool are then partially or fully transferred to theincorporated litter pools, depending on the tillage practice:

Clitter,inc,t+1 = Clitter,inc,t +Clitter,surf,t ·TL for carbon andNlitter,inc,t+1 = Nlitter,inc,t +Nlitter,surf,t ·TL for nitrogen. (1)

The Clitter,surf and Nlitter,surf pools are reduced accordingly:

Clitter,surf,t+1 = Clitter,surf,t · (1−TL),Nlitter,surf,t+1 = Nlitter,surf,t · (1−TL), (2)

where Clitter,inc and Nlitter,inc are the amounts of incorporatedsurface litter C and N, respectively, in grammes per squaremetre (g m−2) at a time step t (days). The parameter TL is thetillage efficiency, which determines the fraction of residuesthat is incorporated by tillage (0–1). To account for the ver-tical displacement of litter through bioturbation under natu-ral vegetation and under no-till conditions, we assume that0.1897 % of the surface litter pool is transferred to the incor-porated litter pool per day (equivalent to an annual bioturba-tion rate of 50 %).

The litter pools are subject to decomposition. The decom-position of litter depends on the temperature and moisture ofits surroundings. The decomposition of the incorporated lit-ter pools depends on soil moisture and temperature of the firstsoil layer (as described by von Bloh et al., 2018), whereasthe decomposition of the surface litter pools depends on thelitter’s moisture and temperature, which are approximatedby the model. The decomposition rate of litter (rdecom ingC m−2 d−1) is described by first-order kinetics, and is spe-cific for each plant functional type (PFT) following Sitch etal. (2003);

rdecom(PFT) = 1− exp(−

1τ10(PFT)

· g (Tsurf) ·F (θ)

), (3)

where τ10 is the mean residence time for litter and F(θ)and g(Tsurf) are response functions of the decay rate to lit-ter moisture and litter temperature (Tsurf), respectively. Theresponse function to litter moisture F(θ) is defined as:

F (θ)= 0.0402− 5.005 · θ3+ 4.269 · θ2+ 0.7189 · θ, (4)

where θ is the volume fraction of litter moisture which de-pends on the water holding capacity of the surface litter(whcsurf), the fraction of surface covered by litter (fsurf),the amount of water intercepted by the surface litter (Isurf)(Sect. 3.3.1), and lost through evaporation Esurf (Sect. 3.3.3).

The temperature function g (Tsurf) describes the influenceof temperature of surface litter on decomposition (von Blohet al., 2018):

g (Tsurf)= exp(

308.56 ·1

66.02−

1(Tsurf+56.02)

), (5)

where Tsurf is the temperature of surface litter (Sect. 3.4).A fixed fraction (70 %) of the decomposed Clitter,surf is

mineralized, i.e. emitted as CO2, whereas the remaining hu-mified C is transferred to the soil C pools, where it is then



subject to the soil decomposition rules as described by vonBloh et al. (2018) and Schaphoff et al. (2018a). The min-eralized N (also 70 % of the decomposed litter) is added tothe NH+4 pool of the first soil layer, where it is subjected tofurther transformations (von Bloh et al., 2018), whereas thehumified organic N (30 % of the decomposed litter) is allo-cated to the different organic soil N pools in the same sharesas the humified C. In order to maintain the desired C : N ratioof 15 within the soil (von Bloh et al., 2018), the mineralizedN is subject to microbial immobilization, i.e. the transforma-tion of mineral N to organic N directly reverting some of theN mineralization in the soil.

The presence of surface litter influences the soil waterfluxes and soil temperature of the soil (see Sect. 3.3 and 3.4),and therefore affects the decomposition of the soil carbonand nitrogen pools, including the transformations of mineralN forms. Nitrogen fluxes such as N2O from nitrification anddenitrification, for instance, are partly driven by soil moisture(von Bloh et al., 2018):

FN2O,nitrification,l = K2 ·Kmax · F1 (Tl) · F1(Wsat,l

)·F (pH) ·NH+4,l for nitrification and

FN2O,denitrification,l = rmx2 · F2(Wsat,l

)· F2

(Tl,Corg

)· NO−3,l for denitrification, (6)

where FN2O,nitrification and FN2O,denitrification are the N2O fluxrelated to nitrification and denitrification, respectively, ingN m−2 d−1 in layer l.K2 is the fraction of nitrified N lost asN2O (K2 = 0.02), Kmax is the maximum nitrification rate ofNH+4 (Kmax = 0.1 d

−1). F1 (Tl) and F1(Wsat,l

)are response

functions of soil temperature and water saturation, respec-tively, that limit the nitrification rate. F (pH) is the functiondescribing the response of nitrification rates to soil pH, andNH+4,l and NO

−

3,l the soil ammonium and nitrate concentra-tions in gN m−2, respectively. F2

(Tl,Corg

)and F2

(Wsat,l

)are reactions for soil temperature, soil carbon, and water sat-uration and rmx2 is the fraction of denitrified N lost as N2O(11 %, the remainder is lost as N2). For a detailed descriptionof the N-related processes implemented in LPJmL, we referthe reader to von Bloh et al. (2018).

3.3 Water fluxes

3.3.1 Litter interception

Precipitation and applied irrigation water in LPJmL5 is par-titioned into interception, transpiration, soil evaporation, soilmoisture, and runoff (Jägermeyr et al., 2015). To account forthe interception and evaporation of water by surface litter, thewater can now also be captured by surface litter through litterinterception (Isurf) and be lost through litter evaporation, sub-sequently infiltrates into the soil and/or forms surface runoff.Litter moisture (θ ) is calculated in the following way:

θt+1 =min(whcsurf− θ(t),Isurf · fsurf). (7)

fsurf is calculated by adapting the equation from Gre-gory (1982) that relates the amount of surface litter (dry mat-ter) per square metre (m2) to the fraction of soil covered:

fsurf = 1− exp(−Am ·OMlitter,surf), (8)

where OMlitter,surf is the total mass of dry matter surface lit-ter in grammes per square metre (g m−2) and Am is the areacovered per mass of crop specific residue (m2 g−1). The to-tal mass of surface litter is calculated assuming a fixed C toorganic matter (OM) ratio of 2.38 (CFOM,litter), based on theassumption that 42 % of the organic matter is C, as suggestedby Brady and Weil (2008):

OMlitter,surf = Clitter,surf ·CFOM,litter, (9)

where Clitter,surf is the amount of C stored in the surface litterpool in grammes of carbon per square metre (gC m−2). Weapply the average value of 0.004 forAm from Gregory (1982)to all materials, neglecting variations in surface litter for dif-ferent materials. whcsurf (mm) is the water holding capac-ity of the surface litter and is calculated by multiplying thelitter mass with a conversion factor of 2× 10−3 mm kg−1

(OMlitter,surf) following Enrique et al. (1999).

3.3.2 Soil infiltration

The presence of surface litter enhances infiltration of pre-cipitation or irrigation water into the soil, as soil crustingis reduced and preferential pathways are affected (Ranaivo-son et al., 2017). In order to account for improved infiltrationwith the presence of surface litter, we follow the approach byJägermeyr et al. (2016), which has been developed for imple-menting in situ water harvesting, e.g. by mulching in LPJmL.The infiltration rate (In in mm d−1) depends on the soil watercontent of the first layer and the infiltration parameter p;

In= prir · p√

1−Wa

Wsat,l=1−Wpwp,l=1, (10)

where prir is the daily precipitation and applied irrigationwater in millimetres, Wa the available soil water content inthe first soil layer, and Wsat,l=1 and Wpwp,l=1 the soil wa-ter contents at saturation and permanent wilting point of thefirst layer in millimetres. By default p = 2, but four differ-ent levels are distinguished (p = 3,4,5,6) by Jägermeyr etal. (2016), in order to account for increased infiltration basedon the management intervention. To account for the effectsof surface litter, we here scale the infiltration parameter pbetween 2 and 6, based on the fraction of surface litter cover(fsurf);

p = 2 · (1+ fsurf · 2). (11)

Surplus water that cannot infiltrate forms surface runoff andenters the river system.



3.3.3 Litter and soil evaporation

Evaporation (Esurf in millimetres) from the surface littercover (fsurf) is calculated in a similar manner as evaporationfrom the first soil layer (Schaphoff et al., 2018a). Evapora-tion depends on the vegetation cover (fv), the radiation en-ergy for the vaporization of water (PET), and the water storedin the surface litter that is available to evaporate (ωsurf) rela-tive to whcsurf. Here, fsurf is also taken into account so thatthe fraction of soil uncovered is subject to soil evaporation asdescribed in Schaphoff et al. (2018a);

Esurf = PET ·α ·max(1− fv,0.05) ·ω2surf · fsurf, (12)ωsurf = θ/whcsurf, (13)

where PET is calculated based on the theory of equilibriumevapotranspiration (Jarvis and McNaughton, 1986) and α theempirically derived Priestley–Taylor coefficient (α = 1.32)(Priestley and Taylor, 1972).

The presence of litter at the soil surface reduces the evapo-ration from the soil (Esoil). Esoil (millimetres) corresponds tothe soil evaporation as described in Schaphoff et al. (2018a),and depends on the available energy for vaporization of waterand the available water in the upper 0.3 m of the soil (ωevap).However, with the implementation of tillage, the fraction offsurf now also influences evaporation, i.e. greater soil cover(fsurf) results in a decrease in Esoil:

Esoil = PET ·α ·max(1− fv,0.05) ·ω2 · (1− fsurf), (14)

where ω is calculated as the evaporation-available water(ωevap) relative to the water holding capacity in that layer(whcevap):

ω =min(

1,ωevap

whcevap

), (15)

where ωevap is all the water above wilting point of the upper0.3 m (Schaphoff et al., 2018a).

3.4 Heat flux

The temperature of the surface litter is calculated as the av-erage of soil temperature of the previous day (t) of the firstlayer (Tsoil,l=1 in ◦C) and actual air temperature (Tair,t+1 in◦C), in the following way:

Tlitter,surf,t+1 = 0.5(Tair,t+1+ Tsoil,l=1,t ). (16)

Equation (16) is an approximate solution for the heat ex-change described by Schaphoff et al. (2013). The new up-per boundary condition (Tupper in ◦C) is now calculated bythe average of Tair and Tsurf weighted by fsurf. With the newboundary condition, the cover of the soil with surface litterdiminishes the heat exchange between soil and atmosphere;

Tupper = Tair · (1− fsurf)+ Tsurf · fsurf. (17)

The remainder of the soil temperature computation remainsunchanged from the description of Schaphoff et al. (2013).

3.5 Tillage effects on physical properties

3.5.1 Dynamic calculation of hydraulic properties

Previous versions of the LPJmL model used static soil hy-draulic parameters as inputs, computed following the pe-dotransfer function (PTF) by Cosby et al. (1984). Differentmethods exist to calculate soil hydraulic properties from soiltexture and SOM content for different points of the water re-tention curve (Balland et al., 2008; Saxton and Rawls, 2006;Wösten et al., 1999) or at continuous pressure levels (VanGenuchten, 1980; Vereecken et al., 2010). Extensive reviewsof PTFs and their application in Earth system and soil mod-elling can be found in Van Looy et al. (2017) and Vereeckenet al. (2016). We now introduce an approach following thePTF by Saxton and Rawls (2006), which was included inthe model in order to dynamically simulate layer-specifichydraulic parameters that account for the amount of SOMin each layer, constituting an important mechanism of howhydraulic parameters are affected by tillage (Strudley et al.,2008).

As such, Saxton and Rawls (2006) define a PTF most suit-able for our needs and capable of calculating all the neces-sary soil water properties for our approach: The PTF allowsfor a dynamic effect of SOM on soil hydraulic properties, andis also capable of representing changes in bulk density aftertillage and was developed from a large number of data points.With this implementation, soil hydraulic properties are nowall updated daily. Following Saxton and Rawls (2006), soilwater properties are calculated as

λpwp,l = −0.024 · a+ 0.0487 ·Cl+ 0.006 ·SOMl + 0.005·Sa ·SOMl − 0.013 ·Cl ·SOMl + 0.068·Sa ·Cl+ 0.031, (18)

Wpwp,l = 1.14 · λpwp,l − 0.02, (19)λfc,l =−0.251 ·Sa+ 0.195 ·Cl+ 0.011 ·SOMl + 0.006·Sa ·SOMl − 0.027 ·Cl ·SOMl + 0.452·Sa ·Cl+ 0.299, (20)

Wfc,l = 1.238 ·(λfc,l

)2+ 0.626 · λfc,l − 0.015, (21)

λsat,l = 0.278 ·Sa+ 0.034 ·Cl+ 0.022 ·SOMl − 0.018·Sa ·SOMl − 0.027 ·Cl ·SOMl − 0.584·Sa ·Cl+ 0.078, (22)

Wsat,l =Wfc,l + 1.636 · λsat,l − 0.097 ·Sa− 0.064, (23)BDsoil,l = (1−Wsat,l) ·MD. (24)

SOMl is the soil organic matter content in weight percent(wt %) of layer l; Wpwp,l is the moisture content at the per-manent wilting point; Wfc,l is moisture contents at field ca-pacity; Wsat,l is the moisture contents at saturation; λpwp,l ,λfc,l , and λsat,l are the moisture contents for the first solutionat permanent wilting point, field capacity, and saturation, re-spectively; Sa is the sand content in volume percent (vol %);



Cl is the clay content in volume percent (vol %); BDsoil,l isthe bulk density in kilogrammes per cubic metre (kg m−3);and MD is the mineral density of 2700 kg m−3. For SOMl ,total SOC content is translated into SOM of this layer:

SOMl =CFOM,soil · (CfastSoil,l +CslowSoil,l)

BDsoil,l · zl· 100, (25)

where CFOM,soil is the conversion factor of 2 as suggestedby Pribyl (2010), assuming that SOM contains 50 % SOC;CfastSoil,l is the fast decaying C pool in kilogrammes persquare metre (kg m−2); CslowSoil,l is the slow decaying Cpool (kg m−2); BDsoil,l is the bulk density in kg m−3; andz is the thickness of layer l in metres. It was suggested bySaxton and Rawls (2006) that the PTF should not be used forSOM contents above 8 %, so we cap SOMl at this maximumwhen computing soil hydraulic properties and thus treatedsoils with SOMl content above this threshold as soils with8 % SOM content. Saturated hydraulic conductivity is alsocalculated following Saxton and Rawls (2006) as

Ksl = 1930 ·(Wsat(l) −Wfc(l)

)3−φl , (26)φl =

ln(Wfc,l

)− ln(Wpwp,l)

ln(1500)− ln(33), (27)

where Ksl is the saturated hydraulic conductivity in millime-tres per hour (mm h−1) and φl is the slope of the logarithmictension–moisture curve of layer l.

3.5.2 Bulk density effect and reconsolidation

The effects of tillage on BD are adopted from the APEXmodel by Williams et al. (2015), which is a follow-up devel-opment of the EPIC model (Williams et al., 1983). Tillagecauses changes in BD of the tillage layer (first topsoil layerof 0.2 m) after tillage. Soil moisture content for the tillagelayer is updated using the fraction of change in BD. Ksl isalso updated based on the new moisture content after tillage.A mixing efficiency parameter (mE), depending on the in-tensity and type of tillage (0–1), determines the fraction ofchange in BD after tillage. A mE of 0.90, for example, rep-resents a full inversion tillage practice, also known as con-ventional tillage (White et al., 2010). The parameter mE canbe used in combination with residue management assump-tions to simulate different tillage types. It should be notedthat Williams et al. (1983) calculated direct effects of tillageon BD, while we changed the equation accordingly to ac-count for the fraction at which BD is changed.

The fraction of BD change after tillage is calculated in thefollowing way:

fBDtill,t+1 = fBDtill,t −(fBDtill,t − 0.667

)·mE. (28)

Tillage density effects on saturation and field capacity followSaxton and Rawls (2006):

Wsat,till,l,t+1 = 1−(1−Wsat,l,t

)· fBDtill,t+1, (29)

Wfc,till,l,t+1 =Wfc,l,t − 0.2 · (Wsat,l,t −Wsat,till,l,t+1), (30)

where fBDtill,t+1 is the fraction of density change of the top-soil layer after tillage, fBDtill,t is the density effect beforetillage, Wsat,till,l,t+1 and Wfc,till,l,t+1 are adjusted moisturecontents at saturation and field capacity after tillage, andWsat,l,t and Wfc,l,t are the moisture content at saturation andfield capacity before tillage.

Reconsolidation of the tilled soil layer is accounted for fol-lowing the same approach by Williams et al. (2015). The rateof reconsolidation depends on the rate of infiltration and thesand content of the soil. This ensures that the porosity andBD changes caused by tillage gradually return to their initialvalue before tillage. Reconsolidation is calculated the follow-ing way:

sz= 0.2 · In ·1+ 2 ·Sa/(Sa+ e8.597−0.075·Sa)

z0.6till, (31)

f =sz

sz+ e3.92−0.0226·sz, (32)

fBDtill,t+1 = fBDtill,t + f · (1− fBDtill,t ), (33)

where sz is the scaling factor for the tillage layer and ztill isthe depth of the tilled layer in metres. This allows for a fastersettling of recently tilled soils with high precipitation and forsoils with a high sand content. In dry areas with low precip-itation, and for soils with a low-sand content, the soil set-tles slower and might not consolidate back to its initial state.This is accounted for by taking the previous bulk densitybefore tillage into account. The effect of tillage on BD canvary from year to year, but fBDtill,t cannot be below 0.667 orabove 1 so that unwanted amplification is not possible. We donot yet account for fluffy soil syndrome processes (i.e. whenthe soil does not settle over time) and negative implicationsfrom this, which results in an unfavourable soil particle dis-tribution that can cause a decline in productivity (Daigh andDeJong-Hughes, 2017).

4 Model set-up

4.1 Model input, initialization, and spin-up

In order to bring vegetation patterns and SOM pools into adynamic equilibrium stage, we make use of a 5000-year spin-up simulation of only natural vegetation, which recycles thefirst 30 years of climate input following the procedures of vonBloh et al. (2018). For simulations with land-use inputs andto account for agricultural management, a second spin-upof 390 years is conducted to account for historical land-usechange, which is introduced in the year 1700. The spatial res-olution of all input data and model simulations is 0.5◦. Land-use data are based on crop-specific shares of MIRCA2000



(Portmann et al., 2010) and cropland and grassland time se-ries since 1700 from HYDE3 (Klein Goldewijk et al., 2010)as described by Fader et al. (2010). As per default setting,intercrops are grown on all set-aside stands in all simula-tions (Bondeau et al., 2007). As we are here interested inthe effects of tillage on cropland, we ignore all natural vege-tation in grid cells with cropland by scaling existing croplandshares to 100 %. We drive the model with daily mean temper-ature from the Climate Research Unit (CRU TS version 3.23;University of East Anglia Climate Research Unit, 2015; Har-ris et al., 2014), monthly precipitation data from the GlobalPrecipitation Climatology Centre (GPCC Full Data Reanal-ysis version 7.0; Becker et al., 2013), and shortwave down-ward and net longwave downward radiation data from theERA-Interim data set (Dee et al., 2011). Static soil textureclasses are taken from the Harmonized World Soil Database(HWSD) version 1.1 (Nachtergaele et al., 2009) and aggre-gated to 0.5◦ resolution by using the dominant soil type.Twelve different soil textural classes are distinguished ac-cording to the USDA soil texture classification and one un-productive soil type, which is referred to as “rock and ice”.Soil pH data are taken from the WISE data set (Batjes, 2005).The NOAA/ESRL Mauna Loa station (Tans and Keeling,2015) provides atmospheric CO2 concentrations. Depositionof N was taken from the ACCMIP database (Lamarque et al.,2013).

4.2 Simulation options and evaluation set-up

The new tillage management implementation allows forspecifying different tillage and residue systems. We con-ducted four contrasting simulations on current cropland areawith or without the application of tillage and with or withoutremoval of residues (Table 1). The default setting for con-ventional tillage is mE= 0.9 and TL= 0.95. In the tillagescenario, tillage is conducted twice a year, at sowing and af-ter harvest. Soil water properties are updated on a daily basis,enabling the tillage effect to be effective from the subsequentday onwards until it wears off due to soil settling processes.The four different management settings (MSs) for globalsimulations are as the following: (1) full tillage and residuesleft on the field (T_R), (2) full tillage and residues are re-moved (T_NR), (3) no-till practice and residues are retainedon the field (NT_R), and (4) no-till practice and residues areremoved from the field (NT_NR). The specific parametersfor these four settings are listed in Table 1. The default MS isT_R and was introduced in the second spin-up from the year1700 onwards, as soon as human land use is introduced in theindividual grid cells (Fader et al. 2010). All of the four MSsimulations were run for 109 years, starting from year 1900.Unless specified differently, the outputs of the four differentMS simulations were analysed using the relative differencesbetween each output variable using T_R as the baseline MS;

RDX =XMS

XT_R− 1, (34)

where RDX is the relative difference between the manage-ment scenarios for variable X and XMS and XT_R are thevalues of variable X of the MS of interest and the baselinemanagement systems: conventional tillage with residues lefton the field (T_R). Spin-up simulations and relative differ-ences for Eq. (34) were adjusted if a different MS was usedas reference system, e.g. if reference data are available forcomparisons of different MSs. The effects were analysed fordifferent time scales: the 3-year average of years 1 to 3 forshort-term effects, the average after years 9 to 11 for mid-term effects, and the average of years 19 to 21 for long-termeffects. Depending on available reference data in the litera-ture, the specific duration and default MS of the experimentwere chosen. The results of the simulations are compared toliterature values from selected meta-analyses. Meta-analysesallow for the comparison of globally modelled results to aset of combined results of individual studies from all aroundthe world, assuming that the data basis presented in meta-analyses is representative. A comparison to individual site-specific studies would require detailed site-specific simula-tions making use of climatic records for that site and detailson the specific land-use history. Results of individual site-specific experiments can differ substantially between sites,which hampers the interpretation at larger scales. We cal-culated the median and the 5th and 95th percentiles (valueswithin brackets) between MS in order to compare the modelresults to the meta-analyses, where averages and 95 % con-fidence intervals (CIs) are mostly reported. We chose me-dians rather than arithmetic averages to reduce outlier ef-fects, which is especially important for relative changes thatstrongly depend on the baseline value. If region-specific val-ues were reported in the meta-analyses, e.g. climate zones,we compared model results of these individual regions, fol-lowing the same approach for each study, to the reported re-gional value ranges.

To analyse the effectiveness of selected individual pro-cesses (see Fig. 1) without confounding feedback processes,we conducted additional simulations of the four differentMSs on bare soil with uniform dry matter litter input (simu-lation NT_NR_bs and NT_R_bs1 to NT_R_bs5) of uniformcomposition (C : N ratio of 20), no atmospheric N deposi-tion and static fertilizer input (Elliott et al., 2015). This helpswith isolating soil processes, as any feedbacks via vegetationperformance are eliminated in this setting.

5 Evaluation and discussion

5.1 Tillage effects on hydraulic properties

Table 2 presents the calculated soil hydraulic properties oftillage for each of the soil classes prior to and after tillage(mE of 0.9), combined with a SOM content in the tilled soillayer of 0 % and 8 %. In general, both tillage and a higherSOM content tend to increase whc, Wsat,l , Wfc,l , and Ksl .



Table 1. LPJmL simulation settings and tillage parameters used in the stylized simulations for model evaluation.

Scenario Simulation Retained residue Tillage efficiency Mixing efficiency of Litter covera Litter amountabbreviation fraction on field (TLFrac) tillage (mE) (%) (dry matter g m2)

Tillage + residues on 100 %scaled cropland

T_R 1 0.95 0.9 variableb variableb

Tillage + no residues on100 % scaled cropland

T_NR 0.1 0.95 0.9 variableb variableb

No-till + residues on 100 %scaled cropland

NT_R 1 0 0 variableb variableb

No-till + no residues on100 % scaled cropland

NT_NR 0.1 0 0 variableb variableb

No-till + no residues on baresoil

NT_NR_bs 0 0 0 0 0

No-till + residues on baresoil (1)

NT_R_bs1 1 0 0 10 17


NT_R_bs2 1 0 0 30 60


NT_R_bs3 1 0 0 50 117


NT_R_bs4 1 0 0 70 202


NT_R_bs5 1 0 0 90 383

a Litter cover is calculated following Gregory (1982). b Litter amounts and litter cover are modelled internally.

Clay soils are an exception, since higher SOM content de-creases whc, Wsat,l , and Wfc,l , and increases Ksl . The effectof increasing SOM content on whc,Wsat,l , andWfc,l is great-est in the soil classes sand and loamy sand. The increasingeffects of tillage on the hydraulic properties are generallyweaker compared to an increase in SOM by 8 % (maximumSOM content for computing soil hydraulic properties in themodel). While tillage (mE of 0.9, 0 % SOM) in sandy soilsincrease whc by 83 %, 8 % of SOM can increase whc in anuntilled soil by 105 % and in a tilled soil by 84 %. As com-parison, in silty loam soils with 0 % SOM, tillage (mE of 0.9)increases whc by 16 %, while 8 % SOM can increase whc by31 % and by 26 % for untilled and tilled soil, respectively.

The PTF by Saxton and Rawls (2006) uses an empiricalrelationship between SOM, soil texture, and hydraulic prop-erties derived from the USDA soil database, implying thatthe PTF is likely to be more accurate within the US than out-side. A PTF developed for global-scale application is, to ourknowledge, not yet developed. Nevertheless PTFs are usedin a variety of global applications, despite the limitations tovalidate at this scale (Van Looy et al., 2017).

5.2 Productivity

In our simulations, adopting NT_R slightly increases produc-tivity for all rain-fed crops simulated (wheat, maize, pulses,rapeseed) on average, but ranges from increases to decreasesacross all cropland globally. This increase can be observedfor the first 3 years (Fig. S2 in the Supplement), and for thefirst 10 years (Fig. 2a and b). All the results shown here andin the subsequent sections are calculated as RD following

Eq. (34), unless otherwise stated. The numbers discussed inthis section refer to the productivity after 10 years (average ofyear 9–11). The largest positive impact can be found for rape-seed, where NT_R results in a median increase of +3.5 %(5th, 95th percentiles: −24.5 %, +57.8 %). The positive im-pact is lowest for maize, with median increases by +1.8 %(5th, 95th percentiles: −24.6 %, +56.2 %). The median pro-ductivity of wheat increases slightly by +2.5 % (5th, 95thpercentiles: −15.2 %, +53.5 %) under NT_R. The slight in-creases in median productivity under NT_R are contrastingto the values reported by Pittelkow et al. (2015b), who re-port slight decreases in productivity for wheat and maize andsmall increases for rapeseed (Table 3). They report both pos-itive and negative effects for wheat and rapeseed, but onlynegative effects for maize. Pittelkow et al. (2015b) identifyaridity and crop type as the most important factors influenc-ing the responses of productivity to the introduction of no-till systems with residues left on the field. The aridity indexwas determined by dividing the mean annual precipitation bypotential evaporation. No-till performed best under rain-fedconditions in dry climates (aridity index < 0.65), by whichthe overall response was equal or positive compared to T_R.

The positive effects on productivity under NT_R in dryregions can also be found in our simulations. For instance,wheat productivity increases substantially under NT_R,whereas this effect diminishes with increases in aridity in-dexes (Fig. 2a). Similar results are found for maize produc-tivity (Fig. 2b). This positive effect can be attributed to thepresence of surface litter, which leads to higher soil moistureconservation through increased water infiltration into the soil



Figure 2. Relative yield changes for rain-fed wheat (a) and rain-fed maize (b) compared to aridity indexes after 10 years NT_R vs. T_R.Low aridity index values indicate arid conditions as the index is defined as mean annual precipitation divided by potential evapotranspiration,following Pittelkow et al. (2015a). Substantial increases in crop yields only occur in arid regions, with aridity indices < 0.75.

and decreases in evaporation. Areas where crop productiv-ity is limited by soil water could therefore potentially ben-efit from NT_R (Pittelkow et al., 2015a). The influence ofclimatic condition of no-till effects on productivity was al-ready found by several other studies (e.g. Ogle et al., 2012;Pittelkow et al., 2015a; van Kessel et al., 2013). Ogle etal. (2012) found declines in productivity but these declineswere larger in the cooler and wetter climates. Pittelkow etal. (2015a) found only small declines in productivity in dry

areas but emphasized that increases in yield can be foundwhen no-till is combined with residues and crop rotation.This was not the case for humid areas (aridity index> 0.65);there declines in productivity were larger under no-till re-gardless of whether residues and crop rotations were applied.Finally, van Kessel et al. (2013) found declines in productiv-ity after adapting to no-till in dry areas (−11 %) and humidareas (−3 %). However, in their analysis it is not clear how



Tabl

e2.

Perc

enta

geva

lues

for

each

soil

text

ural

clas

sof

silt,

sand

and

clay

cont

entu

sed

inL

PJm

Lan

dco

rres

pond

enth

ydra

ulic

para

met

ers

befo

rean

daf

ter

tilla

gew

ith0

%an

d8

%SO

Mus

ing

the

Saxt

onan

dR

awls

(200

6)pe

dotr

ansf

erfu

nctio

n.

Pre-

tilla

ge,0

%SO

Mb

Pre-

tilla

ge,8

%SO

MA

fter

tilla

gec ,

0%

SOM

Aft

ertil

lage

c ,8

%SO

M

Soil

clas

sSi

lt(%

)Sa

nd(%

)C

lay

(%)

whc

dW

sat

Wfc

Ks

whc

Wsa

tW

fcK

sw

hcW

sat

Wfc

Ks

whc

Wsa

tW

fcK

s

Sand

592

30.

040.

420.

0515

2.05

0.09

0.71

0.19

361.

980.

080.

590.

0934

3.67

0.14

0.80

0.21

498.

92L

oam

ysa

nd12

826

0.06

0.40

0.09

83.2

30.

120.

700.

2324

4.20

0.10

0.58

0.13

230.

130.

170.

790.

2536

0.89

Sand

ylo

am32

5810

0.12

0.40

0.17

32.0

30.

180.

700.

3115

2.75

0.15

0.58

0.21

125.

750.

230.

790.

3323

9.93

Loa

m39

4318

0.15

0.41

0.26

10.6

90.

210.

690.

3780

.46

0.19

0.59

0.30

64.7

60.

250.

780.

3914

3.99

Silty

loam

7017

130.

220.

420.

315.

490.

290.

750.

4299

.77

0.26

0.59

0.34

48.2

30.

320.

830.

4415

5.38

Sand

ycl

aylo

am15

5827

0.12

0.42

0.28

6.60

0.17

0.63

0.38

36.3

30.

160.

590.

3248

.79

0.21

0.74

0.40

87.4

0C

lay

loam

3432

340.

170.

470.

382.

290.

200.

650.

4324

.96

0.21

0.63

0.41

26.2

20.

230.

750.

4563

.73

Silty

clay

loam

5610

340.

210.

500.

421.

930.

230.

690.

4534

.54

0.24

0.65

0.45

22.4

50.

250.

780.

4773

.85

Sand

ycl

ay6

5242

0.15

0.47

0.40

0.72

0.16

0.58

0.44

5.64

0.18

0.63

0.44

16.7

30.

200.

700.

4729

.30

Silty

clay

loam

476

470.

200.

560.

481.

640.

180.

650.

4618

.69

0.23

0.69

0.50

16.6

70.

200.

760.

4850

.99

Cla

y20

2258

0.19

0.58

0.53

0.39

0.14

0.58

0.48

2.87

0.21

0.71

0.55

8.62

0.16

0.71

0.50

20.0

3R

ocka

099

10.

000.

010.

010.

100.

000.

010.

010.

100.

000.

010.

010.

100.

000.

010.

010.

10

aSo

ilcl

ass

rock

isno

taff

ecte

dby

SOM

chan

ges

and

tilla

gepr

actic

es.b

ForS

OM

we

only

cons

ider

the

Cpa

rtin

SOM

ingr

amm

esof

carb

onpe

rsqu

are

met

re(g

Cm−

2 ).c

Tilla

gew

itha

mE

of0.

9fo

rcon

vent

iona

ltill

age.

dw

hcis

calc

ulat

edas

whc=W

fc−W

pwp

inal

lcas

es.

crop residues are treated in no-till and tillage (i.e. removed orretained).

Negative effects of NT_R on productivity can be ob-served in mainly tropical areas. As soil moisture increasesin tropical areas under NT_R as well (Fig. 5c), the declineis resulting from a decrease in N availability in the soil(Fig. 5d). Soil moisture drives many N-related processes thatcan cause a decline in N. For instance, the increase in soilmoisture can lead to an increase in denitrification, which de-creases the amount of NO−3 (which will be discussed fur-ther in Sect. 5.5). On the other hand, mineralization can alsobe reduced if soil moisture is too high. However, the soil-moisture–N availability and yield feedback is complex asmany processes are involved.

5.3 Soil C stocks and fluxes

We evaluate the effects of tillage and residue managementon simulated soil C dynamics and fluxes for CO2 emissionsfrom cropland soils, relative change in C input, SOC turnovertime, and relative changes in soil and litter C stocks of thetopsoil (0.3 m). In our simulation CO2 emissions initially de-crease for the average of the first 3 years by a median valueof−11.9 % (5th, 95th percentile:−24.1 %,+2.0 %) after in-troducing no-till (NT_R vs. T_R) (Fig. S3a in the Supple-ment) and soil and litter C stocks increase. After 10 yearsduration (average of year 9–11), however, both CO2 emis-sions and soil and litter C stocks are higher under NT_R thanunder T_R (Fig. 3a, d). Median CO2 emissions from NT_Rcompared to T_R increase by +1.7 % (5th, 95th percentile:−17.4 %,+32.4 %) (Fig. 3a), while at the same time mediantopsoil and litter C also increase by +5.3 % (5th, 95th per-centile: +1.4 %, +12.8 %) (Fig. 3d), i.e. the soil and litterC stock has already increased enough to sustain higher CO2emissions. There are two explanations for CO2 increase inthe long term: (1) more C input from increased net primaryproduction (NPP) for NT_R or (2) a higher decompositionrate over time under NT_R, due to changes in, for exam-ple, soil moisture or temperature. Initially CO2 emissions de-crease almost globally due to increased turnover times underT_R (Fig. S3c in the Supplement), but after 10 years, CO2emissions start to increase in drier regions, while they stilldecrease in most humid regions (Fig. 3a). The median of therelative differences in mean residence time of soil carbon forNT_R compared to T_R is small, but variable (+0.0 % after10 years, 5th, 95th percentile: −22.9 %, +23.7 %) (Fig. 3c),and mean residence time shows similar spatial patterns, i.e.it decreases in drier areas but increases in more humid ar-eas. The drier regions are also the areas where we observea positive effect of reduced evaporation and increased in-filtration on plant growth, i.e. in these regions the C inputinto soils is substantially increased under NT_R compared toT_R (Fig. 3b) (see also Sect. 5.2 for productivity). As such,both mechanisms that affect CO2 emissions are reinforcingeach other in many regions. This is in agreement with the



Table3.

Com

parisonof

simulated

model

outputand

literaturevalues

fromm

eta-analyses.V

aluesfor

modelled

resultsare

calculatedaccording

toE

q.(34)

with

adjusteddefault

managem

ent.

Variable/

SoildepthN

o.ofpairedL

iteraturem

eanTim

ehorizon

Modelled

responseM

odelledresponse

Reference

Scenario(m

)treatm

ents(95

%interval)

(years)(m

edian%

)(5

%and

95%

percentile)

No-tillresidue

–tillage

residue

SOM

(0.3m

)0–0.3

101+

5.0

(+1.0,+

9.2) a,d

10 e+

5.3

+1.4,+

12.8

Abdalla

etal.(2016)C

O2

113−

23.0

(−35.0,−

13.8) a

b−

11.9

−24.1,+

2.0

Abdalla

etal.(2016)N

2 O98

+17.3

(+4.6,+

31.1) a

b+

20.8

−3.6,+

325.5M

eietal.(2018)N

2 O(tropical)

123+

74.1

(+34.8,+

119.9) c,d

b+

15.8

−7.3,+

72.1

Meietal.(2018)

N2 O

(warm

temperate)

62+

17.0

(+6.5,+

29.9) c,d

b+

23.2

+6.0,+

182.3

Meietal.(2018)

N2 O

(cooltemperate)

27−

1.7

(−10.5,+

8.4) c,d

b+

23.5

−0.1,+

664.4

Meietal.(2018)

N2 O

(arid)56

+35.0

(+7.5,+

69.0) a

b+

21.1

−1.8,+

496.3

vanK

esseletal.(2013)N

2 O(hum

id)183

−1.5

(−11.6,+

11.1) a

b+

20.7

−9.1,+

63.8

vanK

esseletal.(2013)Y

ield(w

heat)47

−2.6

(−8.2,+

3.8) a

10 e+

2.5

−15.2,+

53.5

Pittelkowetal.(2015b)

Yield

(maize)

64−

7.6

(−10.1,−

4.3) a

10 e+

1.8

−24.6,+

56.2


Yield

(rapeseed)10

+0.7

(−2.8,+

4.1) a

10 e+

3.5

−24.5,+

57.8


Tillageno

residue–

no-tillnoresidue

SOM

(0.3m

)0–0.3

46−

12.0

(−15.3,−

5.1) a

20 e−

18.0

−42.5,−

0.5

Abdalla

etal.(2016)C

O2

46+

18.0

(+9.4,+

27.3) a

20 e+

21.3

−1.1,+

125.2

Abdalla

etal.(2016)Y

ield(w

heat)B8

+2.7

(−6.3,+

12.7) a

10 e−

5.9

−15.7,+

3.7

Pittelkowatal.(2015b)

Yield

(maize)B

12−

25.4

(−14.7,−

34.1) a

10 e−

5.0

−27.3,+

12.0


tillageno

residue–

tillageresidue

N2 O

105+

1.3

(−5.4,+

8.2) a,d

b−

9.7

−22.0,+

3.6

Meietal.(2018)

aE

stimated

fromgraph. b

Time

horizonofthe

studyis

unclearinthe

meta-analysis.T

heaverage

overthefirst3

yearsofm

odelresultsis

taken. cIncludes

conservationtillage. d

Residue

managem

entforconventionaltillage

unsuree

Time

horizonnotexplicitly

mentioned

byauthor.



meta-analyses conducted by Pittelkow et al. (2015b), whoreport a positive effect on yields (and thus general produc-tivity and thus C input) of no-till compared to conventionaltillage in dry climates. Their results show that, in general,no-till performs best relative to conventional tillage underwater-limited conditions, due to enhanced water-use efficien-cies when residues are retained.

Abdalla et al. (2016) reviewed the effect of tillage, no-till,and residue management and found that if residues are re-turned, no-till compared to conventional tillage increases soiland litter C content by 5.0 % (95th CI: −1.0 %, +9.2 %) anddecreases CO2 emissions from soils by −23.0 % (95th CI:−35.0 %, −13.8 %) (Table 3). These findings of Abdalla etal. (2016) are in line with our findings for CO2 emissionsif we consider the first 3 years of duration for CO2 emis-sions and 10 years duration for topsoil and litter C. Abdallaet al. (2016) do not explicitly specify a time of duration forthese results. If we only analyse the tillage effect withouttaking residues into account (T_NR vs. NT_NR), we find inour simulation that topsoil and litter C decreases by−18.0 %(5th, 95th percentile:−42.5 %,−0.5 %) after 20 years, whileCO2 emissions increase by +21.3 % (5th, 95th percentile:−1.1 %, +125.2 %) mostly in humid regions (Table 3). Ab-dalla et al. (2016) also reported soil and litter C changes froma T_NR vs. NT_NR comparison and reported a decrease insoil and litter C under T_NR of −12.0 % (95th CI: −15.3 %,−5.1 %) and a CO2 increase of +18.0 % (95th CI: +9.4 %,+27.3 %), which is well in line with our model results.

Ogle et al. (2005) conducted a meta-analysis and reportedSOC changes from NT_R compared to T_R system withmedium C input, grouped for different climatic zones. Theyfound a+23 %,+17 %,+16 %, and+10 % mean increase inSOC after converting from a conventional tillage to a no-tillsystem for more than 20 years for tropical moist, tropical dry,temperate moist, and temperate dry climates, respectively.We only find a +4.8 %, +8.3 %, +3.5 %, and +5.8 % meanincrease in topsoil and litter C for these regions, respectively.However, Ogle et al. (2005) analysed the data by comparinga no-till system with high C inputs from rotation and residuesto a conventional tillage system with medium C input fromrotation and residues. We compare two similarly productivesystems with each other, where residues are either left on thefield or incorporated through tillage (NT_R vs. T_R), whichmay explain why we see smaller relative effects in the sim-ulations. Comparing a high input system with a medium ora low input system will essentially lead to an amplificationof soil and litter C changes over time; nevertheless, we arestill able to generally reproduce a SOC increase over longerperiods.

Unfortunately there are high discrepancies in the litera-ture with regard to no-till effects on soil and litter C, sincethe high increases found by Ogle et al. (2005) are not sup-ported by the findings of Abdalla et al. (2016). Ranaivoson etal. (2017) found that crop residues left on the field increases

soil and litter C content, which is in agreement with our sim-ulation results.

5.4 Water fluxes

We evaluate the effects of tillage and residue managementon water fluxes by analysing soil evaporation and surfacerunoff. Our results show that evaporation and surface runoffunder NT_R compared to T_R are generally reduced by−44.3 % (5th, 95th percentiles: −64.5, −17.4 %) and by−57.8 % (5th, 95th percentiles: −74.6 %, −26.1 %), respec-tively (Fig. S4a and b in the Supplement). We also analysedsoil evaporation and surface runoff for different amounts ofsurface litter loads and cover on bare soil without vegetationin order to compare our results to literature estimates fromfield experiments. We find that both the reduction in evapo-ration and surface runoff are dependent on the residue load,which translates into different rates of surface litter cover.

On the process side, water fluxes highly influence plantproductivity and are affected by tillage and residue manage-ment (Fig. 1). Surface litter, which is left on the surface ofthe soil, creates a barrier that reduces evaporation and alsoincreases the rate of infiltration into the soil. Litter that isincorporated into the soil through tillage loses this functionto cover the soil. Both the reduction of soil evaporation andthe increase in rainfall infiltration contribute to increased soilmoisture and hence plant water availability. The model ac-counts for both processes. Scopel et al. (2004) modelled theeffect of maize residues on soil evaporation calibrated fromtwo tropical sites and found that a presence of 100 g m−2 ofsurface litter decreases soil evaporation by −10 % to −15 %in the data, whereas our model shows a median decreasein evaporation of −6.6 % (5th, 95th percentiles: −26.1 %,+20.3 %) globally (Fig. S5a in the Supplement). The effectof a higher amount of surface litter is much more domi-nate as Scopel et al. (2004) found that 600 g m−2 surfacelitter reduced evaporation by approx. −50 %. For the samelitter load, our model shows a median decrease in evapora-tion of −72.6 % (5th, 95th percentiles: −81.5 %, −49.1 %)(Fig. S5b in the Supplement), which is higher than the resultsfound by Scopel et al. (2004). We further analyse and com-pare our model results to the meta-analysis from Ranaivo-son et al. (2017), who reviewed the effect of surface litteron evaporation and surface runoff and other agroecologicalfunctions. Ranaivoson et al. (2017) and the studies compiledby them not explicitly distinguish between the different com-partments of runoff (e.g. lateral-, surface-runoff). We assumethat they measured surface runoff, since lateral runoff is dif-ficult to measure and has to be considered in relation to plotsize. In Fig. 4, modelled global results for relative evapora-tion and surface runoff change for 10 %, 30 %, 50 %, 70 %,and 90 % soil cover on bare soil are compared to litera-ture values from Ranaivoson et al. (2017). Concerning theeffect of soil cover on evaporation (Fig. 4a), we find thatwe are well in line with literature estimates from Ranaivo-



Figure 3. Relative C dynamics for NT_R vs. T_R comparison after 10 years of simulation experiment (average of year 9–11) for relativeCO2 change (a), relative C input change (b), relative change of soil C turnover time (c), and relative topsoil and litter C change (d).

son et al. (2017) for up to 70 % soil cover, especially whenanalysing humid climates. For higher soil cover ≥ 70 %, themodel seems to be more in line with literature values forarid regions. Overall for high soil cover of 90 %, the modelseems to overestimate the reduction of evaporation. It shouldbe noted that the estimates from Ranaivoson et al. (2017)are only taken from two field studies, which are only rep-resentative for the local climatic and soil conditions, sinceglobal data on the effect of surface litter on evaporation arenot available. The general effect of surface litter on the re-duction in soil evaporation is thus captured by the model, butthe model seems to overestimate the response at high litterloads. It is not entirely clear from the literature if these ex-periments have been carried out on bare soil without vegeta-tion. If crops are also grown in the experiments, water can beused for transpiration which is otherwise available for evap-oration, which could explain why the model overestimatesthe effect of surface litter on evaporation on bare soil withoutany vegetation.

Ranaivoson et al. (2017) also investigated the runoff re-duction under soil cover, but the results do not show a clearpicture. In theory, surface litter reduces surface runoff and theliterature generally supports this assumption (Kurothe et al.,2014; Wilson et al., 2008), but the magnitude of the effectvaries. Figure 4b compares our modelled results under dif-ferent soil cover to the literature values from Ranaivoson etal. (2017). This shows that modelled results across all global

cropland are on the upper end of the effect of surface runoffreduction from soil cover, but they are still well within therange reported by Ranaivoson et al. (2017). The amount ofwater which is infiltrated (and thus not going into surfacerunoff) is affected by the parameter p in Eq. (11), which isdependent on the amount of surface litter cover (fsurf). Theparameterization of p is chosen to be at the upper end ofthe approach by Jägermeyr et al. (2016) at full surface lit-ter cover, as this should substantially reduce surface runoff(Tapia-Vargas et al., 2001) and thus increase infiltration rates(Strudley et al., 2008). The parametrization of p can be ad-justed if better site-specific information on slope, soils crust-ing, and rainfall intensity is available.

5.5 N2O fluxes

Switching from tillage to no-till management with leavingresidues on the fields (NT_R vs. T_R) increases N2O emis-sions by a median of +20.8 % (5th, 95th percentile: −3.6 %,+325.5 %) (Fig. S6a in the Supplement). The strongest in-crease is found in the cool temperate zone where the av-erage increase is +23.5 % (5th, 95th percentile: −0.1 %,+664.4 %) (Fig. S6e in the Supplement). The lowest increaseis found in the tropical zone +15.8 % (5th, 95th percentile:−7.3 %, +72.1 %) (Fig. S6c in the Supplement).

The increase in N2O emissions after switching to no-tillis in agreement with several literature studies (Linn and Do-



Figure 4. Relative change in evaporation (a) and surface runoff (b) relative to soil cover from surface residues for different soil cover valuesof 10 %, 30 %, 50 %, 70 %, and 90 % (simulation NT_R_bs1 to NT_R_bs5 vs. NT_NR_bs, respectively). For better visibility, the red andblue boxplots are plotted next to the overall boxplots, but correspond to the soil cover value of the overall simulation (empty boxes).

ran, 1984; Mei et al., 2018; van Kessel et al., 2013; Zhaoet al., 2016) (Table 3). Mei et al. (2018) reports an over-all increase of +17.3 % (95th CI: +4.6 %, +31.1 %), whichis in agreement with our median estimate. However, the re-gional patterns over the different climatic regimes are in lessagreement. LPJmL simulations strongly underestimate theincrease in N2O emissions in the tropical zone, whereas sim-ulations overestimate the response in cool temperate and hu-mid zones and to some extent in the warm temperate zone(Table 3).

In general, N2O emissions are formed in two separate pro-cesses: nitrification and denitrification. The increase in N2Oemissions after adapting to NT_R is mainly resulting fromdenitrification in our simulations (+55.9 %, Fig. 5a). Thisincrease is visible in most of the regions. The N2O emis-sions resulting from nitrification decrease mostly (median of−6.0 %, Fig. 5b) but tends to increase in dry areas. The in-crease in denitrification and decrease in nitrification, resultsin a decrease in NO−3 (median of−26.4 %), which appears tobe stronger in the tropical areas as well (Fig. 5d). The trans-formation of mineral N to N2O is not only affected by thenitrification and denitrification rates, but also by substrateavailability (NH+4 and NO

−

3 ). These in turn are affected bynitrification and denitrification rates, but also by other pro-cesses, such as plant uptake and leaching. In the Sahel zone,for example, denitrification decreases and nitrification in-creases, but NO−3 stocks decline, because leaching increasesmore strongly (Fig. S7 in the Supplement).

In LPJmL, denitrification and nitrification rates are mainlydriven by soil moisture and to a lesser extent by soil tem-perature, soil C (denitrification), and soil pH (nitrification).A strong increase in annually averaged soil moisture can beobserved after adapting NT_R (median of+18.9 %, Fig. 5c).Denitrification, as an anoxic process, increases non-linearly

beyond a soil moisture threshold (von Bloh et al. 2018),whereas there is an optimum soil moisture for nitrification,which is reduced at low and high soil moisture contents. Inwet regions, as in the tropical and humid areas, nitrification isthus reduced by no-till practices, whereas it increases in dryerregions. The increase in soil moisture under NT_R is causedby higher water infiltration rates and reduced soil evapora-tion (see Sect. 5.4). Also, no-till practices tend to increasebulk density and thus higher relative soil moisture contents(Fig. 1) also affecting nitrification and denitrification ratesand therefore N2O emissions (van Kessel et al., 2013; Linnand Doran, 1984).

Empirical evidence shows that the introduction of no-tillpractices on N2O emissions can cause both increases and de-creases in N2O emissions (van Kessel et al., 2013). This vari-ation in response is not surprising, as tillage affects severalbiophysical factors that influence N2O emissions (Fig. 1) inpossibly contrasting manners (van Kessel et al., 2013; Snyderet al., 2009). For instance, no-till can lower soil temperatureexchange between soil and atmosphere through the presenceof litter residues, which can reduce N2O emissions (Enriqueet al., 1999). Reduced N2O emissions under no-till comparedto tillage MS can also be observed in the model results, forinstance in northern Europe and areas in Brazil (Fig. S6a inthe Supplement).

As several biophysical factors are affected, N2O emissionsare characterized by significant spatial and temporal variabil-ity. As a result, the estimation of N2O emissions are accom-panied with high uncertainties (Butterbach-Bahl et al., 2013),which hamper the evaluation of the model results (Chatskikhet al., 2008; Mangalassery et al., 2015).

The deviations from the model results compared to themeta-analyses especially for specific climatic regimes (i.e.tropical and cool temperate) require further investigations



Figure 5. Relative changes for the average of the first 3 years of NT_R vs. T_R for denitrification (a), nitrification (b), soil water content (c),and NO−3 (d).

and verification, including model simulations for specificsites at which experiments have been conducted. The sen-sitivity of N2O emissions highlights the importance of cor-rectly simulating soil moisture. However, simulating soilmoisture is subject to strong feedback with vegetation per-formance and comes with uncertainties, as addressed by, forexample, Seneviratne et al. (2010). The effects of differentmanagement settings (as conducted here) on N2O emissionsand soil moisture therefore requires further analyses, ideallyin different climate regimes, soil types, and in combinationwith other management settings (e.g. N fertilizers). We ex-pect that further studies using this tillage implementation inLPJmL will increase the understanding of management ef-fects on soil nitrogen dynamics. The great diversity in ob-served responses in N2O emissions to management options(Mei et al., 2018) renders modelling these effects as chal-lenging, but we trust that the ability of LPJmL5.0-tillage torepresent the different components can also help to better un-derstand their interaction under different environmental con-ditions.

5.6 General discussion

The implementation of tillage into the global ecosystemmodel LPJmL opens up opportunities to assess the effectsof different tillage practices on agricultural productivity andits environmental impacts, such as nutrient cycles, water con-

sumption, GHG emissions, and C sequestration and is a gen-eral model improvement to the previous version of LPJmL(von Bloh et al., 2018). The implementation involved (1) theintroduction of a surface litter pool that is incorporated intothe soil column at tillage events and the subsequent effects onsoil evaporation and infiltration, (2) dynamically accountingfor SOM content in computing soil hydraulic properties, and(3) simulating tillage effects on bulk density and the subse-quent effects of changed soil water properties and all water-dependent processes (Fig. 1).

In general, a global model implementation on tillage prac-tices is difficult to evaluate as effects are often reported tobe quite variable, depending on local soil and climatic con-ditions. The model results were evaluated with data com-piled from meta-analyses, which implies several limitations.Due to the limited amount of available meta-analyses, notall fluxes and stocks could be evaluated within the differ-ent management scenarios. For the evaluation we focused onproductivity, soil and litter C stocks and fluxes, water fluxes,and N2O dynamics. The sample size in some of these meta-analyses was sometimes low, which may result in biases ifan unrepresentative set of climate and soil combinations wastested. Clearly a comparison of a small sample size to sim-ulations of the global cropland is challenging. Nevertheless,the meta-analyses gave the best overview of the overall ef-fects of tillage practices that have been reported for variousindividual experiments.



We find that the model results for NT_R compared to T_Rare generally in agreement with literature with regard to mag-nitude and direction of the effects on C stocks and fluxes. De-spite some disagreement between reported ranges in effectsand model simulations, we find that the diversity in mod-elled responses across environmental gradients is an asset ofthe model. The underlying model mechanisms, such as theinitial decrease in CO2 emissions after introduction of no-tillpractices, can be maintained for longer time periods in moistregions, but are inverted in dry regions due to the feedbackof higher water availability on plant productivity and reducedturnover times, and generally increasing soil carbon stocks(Fig. 3) are plausible and in line with general process under-standing. Certainly, the interaction of the different processesmay not be captured correctly and further research on this isneeded. We trust that this model implementation represent-ing this complexity allows for further research in this direc-tion. For water fluxes, the model seems to overestimate theeffect of surface residue cover on evaporation for high sur-face cover, but the evaluation is also constrained by the smallnumber of suitable field studies. Effects can also change overtime so that a comparison needs to consider the timing, his-tory, and duration of management changes and specific lo-cal climatic and soil conditions. The overall effect of NT_Rcompared to T_R on N2O emissions is in agreement with lit-erature as well. However, the regional patterns over the dif-ferent climatic regimes are in less agreement. N2O emissionsare highly variable in space and time and are very sensitiveto soil water dynamics (Butterbach-Bahl et al., 2013). Thesimulation of soil water dynamics differs per soil type as thecalculation of the hydraulic parameters is texture specific.Moreover, these parameters are now changed after a tillageevent. The effects of tillage on N2O emissions, as well asother processes that are driven by soil water (e.g. CO2, waterdynamics), can therefore be different per soil type. The soil-specific effects of tillage on N2O and CO2 emissions was al-ready studied by Abdalla et al. (2016) and Mei et al. (2018).Abdalla et al. (2016) found that differences in CO2 emissionsbetween tilled and untilled soils are largest in sandy soils(+29 %), whereas the differences in clayey soils are muchsmaller (+12 %). Mei et al. (2018) found that clay content< 20 % significantly increases N2O emissions (+42.9 %) af-ter adapting to conservation tillage, whereas this effect forclay content> 20 % is smaller (+2.9 %). These studies showthat soil-type-specific tillage effects on several processes canbe of importance and should be investigated in more detailin future studies. The interaction of all relevant processes iscomplex, as seen in Fig. 1, which can also lead to high un-certainties in the model. Again, we think that this model im-plementation captures substantial aspects of this complexityand thus lays the foundation for further research.

It is important to note that not all processes related totillage and no-till are taken into account in the current modelimplementation. For instance, NT_R can improve soil struc-ture (e.g. aggregates) due to increased faunal activity (Mar-

tins et al., 2009), which can result in a decrease in BD. Al-though tillage can have several advantages for the farmer, e.g.residue incorporation and topsoil loosening, it can also haveseveral disadvantages. For instance, tillage can cause com-paction of the subsoil (Bertolino et al., 2010), which resultsin an increase in BD (Podder et al., 2012) and creates a bar-rier for percolating water, leading to ponding and an oversat-urated topsoil. Strudley et al. (2008), however, observed di-verging effects of tillage and no-till on hydraulic properties,such as BD, Ks, and whc for different locations. They ar-gue that affected processes of agricultural management havecomplex coupled effects on soil hydraulic properties; also,that variations in space and time often lead to higher differ-ences than the measured differences between the manage-ment treatments. They further argue that characteristics ofsoil type and climate are unique for each location, whichcannot simply be transferred from one field location to an-other. A process-based representation of tillage effects as inthis extension of LPJmL allows for further studying of man-agement effects across diverse environmental conditions, butalso to refine model parameters and implementations whereexperimental evidence suggests disagreement.

One of the primary reasons for tillage, weed control, is alsonot accounted for in LPJmL5.0-tillage or in other ecosystemmodels. As such, different tillage and residue managementstrategies can only be assessed with respect to their biogeo-chemical effects, but only partly with respect to their effectson productivity and not with respect to some environmen-tal effects (e.g. pesticide use). Our model simulations showthat crop yields increase under no-till practices in dry areasbut decrease in wetter regions (Fig. 2). However, the medianresponse is positive, which may be in part because the wa-ter saving effects from increased soil cover with residues areoverestimated or because detrimental effects, such as compe-tition with weeds, are not accounted for.

The included processes now allow us to analyse long-termfeedbacks of productivity on soil and litter C stocks and Ndynamics. Nevertheless the results need to be interpretedcarefully, due to the capacity of the model and implementedprocesses. We also find that the modelled impacts of tillageare very diverse in space as a result of different framingconditions (soil, climate, management) and feedback mech-anisms, such as improved productivity in dry areas if residuecover increases plant-available water. In LPJmL5.0-tillage,the process-based representation of tillage and residue man-agement, and the effects on water fluxes such as evaporationand infiltration at the global scale, is unique in the context ofglobal biophysical models (e.g. Friend et al., 2014; LeQuéréet al., 2018). Future research on improved parameterizationand the implementation of a more detailed representation oftillage processes, the effects on soil water processes, changesin porosity and subsoil compaction, and effects on biodiver-sity and soil N dynamics are needed in order to better as-sess the impacts of tillage and residue management at theglobal scale. The spatial resolution needed to resolve pro-



cesses, such as erosion, data availability, and model structure,need to be considered in further model development (Lutz etal., 2019). As such, some processes, such as a detailed repre-sentation of soil crusting processes, may remain out of reachfor global-scale modelling.

6 Conclusions

We described the implementation of tillage-related processesinto the global ecosystem model LPJmL5.0-tillage. The ex-tended model was tested under different management sce-narios and evaluated by comparing to reported impact rangesfrom meta-analyses on C, water, and N dynamics, as well ason crop yields.

We find that mostly arid regions benefit from a no-till man-agement with leaving residues on the field, due to the watersaving effects of surface litter. We are able to broadly repro-duce reported tillage effects on global stocks and fluxes, aswell as regional patterns of these changes with LPJmL5.0-tillage, but deviations in N fluxes need to be further exam-ined. Not all effects of tillage – including one of its pri-mary reasons, weed control – could be accounted for in thisimplementation. Uncertainties mainly arise because of themultiple feedback mechanisms affecting the overall responseto tillage, especially as most processes are affected by soilmoisture. The processes and feedbacks presented in this im-plementation are complex and evaluation of effects is oftenlimited to the availability of reference data. Nonetheless, theimplementation of more detailed tillage-related mechanicsinto the LPJmL global ecosystem model improves our abilityto represent different agricultural systems and to understandmanagement options for climate change adaptation, agricul-tural mitigation of GHG emissions, and sustainable intensifi-cation. We trust that this model implementation and the pub-lication of the underlying source code will promote researchon the role of tillage for agricultural production, its environ-mental impact, and global biogeochemical cycles.

Code and data availability. The source code is publicly avail-able under the GNU AGPL version 3 license. An exact ver-sion of the source code described here is archived underhttps://doi.org/10.5281/zenodo.2652136 (Herzfeld et al., 2019).

Supplement. The supplement related to this article is availableonline at: https://doi.org/10.5194/gmd-12-2419-2019-supplement.

Author contributions. FL and TH both share the lead authorship forthis paper. They had equal input in designing and conducting themodel implementation, model runs, analysis, and writing of the pa-per. SR contributed to simulation analysis and paper preparation andevaluation. JH contributed to the code implementation, evaluation,and analysis, and edited the paper. SS contributed to the code im-

plementation and evaluation and edited the paper. WvB contributedto the code implementation and evaluation and edited the paper. JJScontributed to the study design and edited the paper. CM contributedto the study design and supervised the implementation, simulations,and analyses, and also edited the paper.

Competing interests. The authors declare that they have no conflictof interest.

Acknowledgements. Femke Lutz, Tobias Herzfeld, and SusanneRolinski gratefully acknowledge the German Ministry for Educa-tion and Research (BMBF) for funding this work, which is partof the MACMIT project (01LN1317A). Jens Heinke acknowledgesBMBF funding through the SUSTAg project (031B0170A). Wethank Quazi Rasool and one anonymous referee for their helpfulcomments on earlier versions of the paper.

Financial support. This research has been supported by theBMBF (grant no. 01LN1317A) and through SUSTAg (grant no.031B0170A).

The article processing charges for this open-access publica-tion were covered by the Potsdam Institute for Climate ImpactResearch (PIK).

Review statement. This paper was edited by Havala Pye and re-viewed by Quazi Rasool and one anonymous referee.

References

Abdalla, K., Chivenge, P., Ciais, P., and Chaplot, V.: No-tillagelessens soil CO2 emissions the most under arid and sandy soilconditions: results from a meta-analysis, Biogeosciences, 13,3619–3633, https://doi.org/10.5194/bg-13-3619-2016, 2016.

Armand, R., Bockstaller, C., Auzet, A.-V., and Van Dijk, P.: Runoffgeneration related to intra-field soil surface characteristics vari-ability: Application to conservation tillage context, Soil Till.Res., 102, 27–37, https://doi.org/10.1016/j.still.2008.07.009,2009.

Aslam, T., Choudhary, M. A., and Saggar, S.: Influence ofland-use management on CO2 emissions from a silt loamsoil in New Zealand, Agr. Ecosyst. Environ., 77, 257–262,https://doi.org/10.1016/S0167-8809(99)00102-4, 2000.

Balland, V., Pollacco, J. A. P., and Arp, P. A.: Mod-eling soil hydraulic properties for a wide rangeof soil conditions, Ecol. Model., 219, 300–316,https://doi.org/10.1016/j.ecolmodel.2008.07.009, 2008.

Batjes, N.: ISRIC-WISE global data set of derived soil propertieson a 0.5 by 0.5 degree grid (version 3.0), ISRIC – World SoilInformation, Wageningen, 2005.

Becker, A., Finger, P., Meyer-Christoffer, A., Rudolf, B., Schamm,K., Schneider, U., and Ziese, M.: A description of the globalland-surface precipitation data products of the Global Precipita-


Documents

Simulating the effect of tillage practices with the global ......(Levis et al., 2014; Oleson et al., 2010) PROMET (Mauser and Bach, 2009), and the Dynamic Land Ecosystem Model (DLEM;